Increasing confidence in model prediction: A case study on water quality data collation for model validation in the Great Barrier Reef catchments

Reliable water quality data is critical for the calibration and validation of catchment models. Recent application of the Source Catchments water quality model across the Great Barrier Reef (GBR) catchments of north eastern Australia has highlighted the need for water quality data, especially during high flows, to reduce uncertainty and "reliance on best guess" approaches to model parameterisation and the subsequent validation of model outputs. Under the Reef Rescue program the Queensland Department of Environment and Resource Management (DERM) is required to build six catchment models to report on progress against water quality targets. A modified version of the eWater Source Catchments water quality model is being applied across the GBR region (similar to 423, 000km(2)). A common problem for water quality modelling is a lack of data to parameterise and validate models. It is therefore important to identify and utilise data from a range of sources. If resources are available, there is also the opportunity to acquire new data to inform the modelling. This paper describes the process used for model validation across an extensive geographical area using a diverse source of data sets. Validation data sets included the collation of historical data from the DERM ambient and event monitoring programs, recent high frequency event load monitoring data and previous modelled estimates to compare with current modelled outputs. Additional data sets used in the validation process included continuous turbidity logging data, cross sectional event sampling in combination with acoustic Doppler backscatter data, sediment sourcing to identify contributing areas in the landscape, storage trapping efficiency and concentration half lives for nutrients and pesticides in storages. Collaboration and data sharing were, and continue to be, a major part of the overall data collection and validation process. The use of analysis software such as the "Loads Tool" and the "River Analysis Package" simplified many calculations in a consistent way for comparison to model outputs. The results from this data collation and validation process were highly variable within individual GBR catchments and across the entire GBR region. This variability in water quality and catchment condition data is not surprising given the range of geographical and climatic extremes experienced in north eastern Australia. The amount and reliability of the collated data also varied between regions; however, the combination of collating existing data and collecting new data specific to model parameterisation provided a significant improvement to the knowledge base for current and future model validation. The data validation exercise has provided a much greater understanding of what data is available, how it can be used in the current model applications for the GBR and what new data will be required in the future to increase the confidence in model outputs.